The whole point of coming to Copenhagen was to work with colleagues in the SAXO Institut (for Greek and Latin) learning how to do palaeography. I'm working on an MS from the mid 13th C, an anonymous text on sophismata, and I'm transcribing the one on Tantum verum opponitur falso. My training has essentially been to work through the manuscript during the day, making my stumbling transcription, and then meeting for an hour with someone who has been doing this longer than I've been alive to go through it all to correct all my errors and fill in my holes. Last week things went pretty slowly, as I started learning the idiosyncracies of the text I'm working with, and as I spent more time than planned going to and from, and sitting at, the police station. But, I've now done about three columns, which equals about four A4 LaTeX document class "article" pages.
I'm having a blast with the sleuthing aspect of paleography, but what has independently fascinated me is the information-conveyance aspect of it. The text is heavily abbreviated -- to the point where when I find a word that is written out in full, it often causes me to stumble, because I just don't expect it, especially if it's a short word (for example, I recently found hoc spelled out. I second-guessed myself on it because surely he wouldn't write that out, he'd abbreviate it!) -- which means that a lot of information can be compressed into a small amount of space. However, on the flip side, there's a lot of redundancies: Often the same sentiment is expressed two or three times over, in just minor variants of grammatical structure. You might wonder "what's the point in abbreviating everything and then saying it three times over", instead of just writing it out in full once -- it would probably take about the same amount of space, and hence about the same amount of effort to write -- but there's actually a good reason for repeating things in triplicate, and that is that it makes the text really quite robustly error-proof. If a scribe makes a mistake in copying, whether it's something minor like forgetting a word, or something more major like skipping over a line, if you've got the same information presented in two or three different ways, it's likely that one or more of the other presentations will still be correct, and so the reader can mentally amend the text and correct the scribe's mistake (and in fact this is often what an editor of a text will do). So what you've got is a system of information presentation which is extremely compressed, so you can fit a lot of information in a small amount of space, and yet is relatively error-proof, so that as it is copied, the content of the text is quite stable.
I find this fascinating.